Online Safety Bill screening measures amount to ‘prior restraint’

Specific issues

The opinion outlined a number of specific issues with the bill, including that forcing tech companies to make their own determinations over what is and is not illegal content introduces the risk that “a significant number of lawful posts will be censored” as a result.

It added this is especially problematic given the highly likely use of AI algorithms and automated software to filter through content at scale.

Highlighting the example of existing immigration offences being incorporated into the list of illegal content that tech platforms must proactively police – which means firms could be forced to remove videos of people crossing the English Channel “which show that activity in a positive light” – the opinion noted it is difficult to predict how an AI algorithm will decide if the content is “positive” or not.

“What if the image was posted without any text? Removing these images prior to publication, without due process to decide whether they are illegal, may have a chilling effect on a debate of public importance relating to the treatment of refugees and illegal immigration,” it said, adding that the threat of large fines and criminal sanctions for managers over failures to effectively police illegal content will also incentivise companies to err on the side of caution and block more content than they otherwise would.

It further added that there is a significant and growing body of research that demonstrates AI and other  automated technologies frequently contain “inherent biases”, creating a clear risk that “screening systems will disproportionately block content relating to or posted by minority ethnic or religious groups”.

This will be compounded by an overall lack of transparency and accountability, as the bill currently places no requirements on companies to provide users with a reason for why their content was removed, or to even notify them that is has happened.

“By contrast to the draconian enforcement conditions to encourage prior restraint, the provisions which enable users to challenge over-zealous action…are limited,” it said, adding that while companies will have a duty to operate complaints procedures, there is no information in the bill about the timescales with which these complaints should be addressed, and no enforcement processes for failures to adequately deal with complaints.

A previous legal opinion commissioned by Index on Censorship from Matthew Ryder KC, published in November 2022, found that technical notices issued by Ofcom – which would require private messaging services to put in place “accredited” tech to filter the content of messages – amount to state-mandated surveillance on a mass scale.

“Ofcom will have a wider remit on mass surveillance powers of UK citizens than the UK’s spy agencies, such as GCHQ (under the Investigatory Powers Act 2016),” wrote Ryder.

The surveillance powers proposed by the Online Safety Bill were unlikely to be in accordance with the law and would be open to legal challenge, he said: “Currently, this level of state surveillance would only be possible under the Investigatory Powers Act if there is a threat to national security.”

A number of technology companies offering encrypted messaging services – including WhatsApp, Signal and Threema – have also previously urged the government to make urgent changes to the bill, arguing that it threatens to undermine the privacy of encrypted communications.

The Home Office maintains that that the Online Safety Bill does not represent a ban on end-to-end encryption and would not require messaging services to weaken their encryption.